Traditionally, surveillance and security control have relied on static methods and conventional technologies, which may overlook subtle signs of tension or discomfort in individuals. Consequently, the analysis of human behavior in nervous situations and its implications for security has become a relevant and emerging area of study. In institutional settings, the early detection of unusual behaviors is essential to prevent risk situations. This paper presents the development and evaluation of a real-time monitoring system based on Kinect sensors and supervised learning algorithms aimed at identifying body postures associated with nervousness. The proposed solution incorporates a multilayered architecture and a web-based interface that enables the visualization, recording, and management of alerts related to suspicious behavior. Two evaluation scenarios were tested: a controlled environment, where the system achieved effectiveness 100%, and a real world context, with an accuracy of 58.59%. Furthermore, the SUS usability survey reported an average score of 89, indicating strong user acceptance. The results support the feasibility of the system as a supplementary tool for institutional security, with potential for improvement through the incorporation of more diverse data, continuous learning techniques, and additional sensors suited for complex environments.